PROCEEDINGS of the 3rd Joint Federal Interagency Conference on Sedimentation and Hydrologic Modeling, April 19-23, 2015, Reno, Nevada, USA COLORADO FRONT RANGE FLOOD OF 2013: PEAK FLOWS AND FLOOD FREQUENCIES Steven E. Yochum, Hydrologist, US Forest Service, National Stream & Aquatic Ecology Center, Fort Collins, CO, [email protected] prepared for the 5th Federal Interagency Hydrologic Modeling Conference April 19-23, 2015, Peppermill Hotel, Reno, Nevada, USA Abstract: In September of 2013, the Colorado Front Range foothills experienced an extensive period of rainfall that culminated in a major flood that peaked in many streams on Friday, the 13th. Rainfall depths of up to 18 inches were recorded over a 10 day period, with a large proportion of the rainfall falling over a 36 hour period. These foothill locations on average receive between 17 and 19 inches of precipitation annually; this event delivered an average year of rainfall at some locations. In response, many streams in the South Platte and Arkansas River basins flooded. To quantify the magnitude of the flood peaks, several entities implemented forensic hydrology methods to develop peak flow estimates, including the NRCS, USGS, and retired USGS hydrologist Bob Jarrett. Peak discharges of up to 60,000 cfs were quantified. Peak flow unit discharges varied by catchment size, as would be expected. Unit discharges as large as 1340 cfs/mi2 were measured. For locations with streamgages, revised flow frequency estimates were developed using the logPearson methodology as presented in Bulletin 17B. The 2013 peaks were included in this analysis. For the larger streams impacted by the flooding, this flood had return intervals ranging from a 5- to 25-year flood (Fountain Creek), 25- to 50-year flood (Cache la Poudre River, South Platte River), 100-year flood (Big Thompson River), 100- to 200-year flood (Boulder Creek, Coal Creek), and greater than the 200-year flood (Lefthand Creek, Saint Vrain Creek, Fish Creek). INTRODUCTION In September of 2013, large portions of the Colorado Front Range foothills (Figure 1) received an unusual amount of rainfall, with up to 18 inches falling in 10 days. Raingage data over the most severely-impacted foothills indicate up to 15 inches fell in Larimer County, 18 inches fell in Boulder County, and 16 inches fell in El Paso County, the three counties most impacted by the flooding. The highest measured rainfall depths are similar to the average annual rainfall for these areas. The majority of the precipitation fell during 36 hours, on September 11th and 12th. These rainfall data were collected in settled areas and primarily within valley bottoms in many portions of the flood extent; rainfall depths and intensities may have been even greater on some mountain slopes (driven by orographic lift) and in remote areas that were void of ground-based data collection. As a result, large floods occurred in the South Platte and Arkansas basins, in the Cache la Poudre, Little Thompson, Big Thompson, and South Platte Rivers, and in the Saint Vrain, Left Hand, Boulder, Coal, and Fountain Creeks (Figure 1). Peak flow estimates were developed using forensic hydrology methodologies in these and their contributing streams. Using these peak flow estimates, revised flood discharge relationships were developed at streamgage locations. This report provides a summary of peak flow estimates developed primarily by the Natural Resources 3rdJFIC, 2015 537 10thFISC+5thFIHMC PROCEEDINGS of the 3rd Joint Federal Interagency Conference on Sedimentation and Hydrologic Modeling, April 19-23, 2015, Reno, Nevada, USA Figure 1 Rainfall depths and streams impacted by the 2013 Colorado Front Range Flood. 3rdJFIC, 2015 538 10thFISC+5thFIHMC PROCEEDINGS of the 3rd Joint Federal Interagency Conference on Sedimentation and Hydrologic Modeling, April 19-23, 2015, Reno, Nevada, USA Conservation Service (NRCS), the U.S. Geological Survey (USGS) and by Bob Jarrett, a retired USGS research hydrologist. Unit discharges are also presented. Based on updated streamgage analyses, return intervals of the flood peaks are provided. METHODS Peak Flow Estimation Data collection efforts were performed by the NRCS (Steven Yochum and Dan Moore), the USGS, and Bob Jarrett. Peak flow estimates were developed using several methods, including the slope-conveyance, step-backwater, and critical depth methods. Two-dimensional computational modeling was also performed for one site (St. Vrain Creek in Lyons). The slope- conveyance method assumes uniform flow and applies the Manning’s equation to cross sectional data. This method requires a Manning’s n estimate, which is problematic since n varies by stage and debris loading (Limerinos, 1970; Bathurst, 1985; Lee and Ferguson, 2002; Wilcox and Wohl, 2006; Reid and Hickin, 2008; Ferguson, 2010; Yochum et al. 2012), and since little field- based data has been collected during large floods for informing the computations. The step- backwater method implements a series of cross sections and 1-dimensional gradually-varied flow computations using such software as HEC-RAS. Discharge is varied until a series of high water marks are matched by the simulated water surface. This method also relies upon uncertain Manning’s n estimates. The critical depth method assumes critical flow at a control cross section, such as at a constriction or drop off of a channel, or assumes critical flow in a high gradient channel. For slopes over 1 percent, it has been found that a critical depth assumption, with 3 to 6 replicate estimates from multiple cross sections within the reach of interest, can obtain estimates within ± %15 of discharges measured with current meters (Jarrett and Tomlinson 2000; Webb and Jarrett 2002; Jarrett and England 2002). This method avoids estimation errors imposed by unknown Manning’s n coefficients for large flood events, but is instead reliant upon the critical depth assumption and is sensitive to such conditions as backwater and localized supercritical flow. A more detailed discussion of each method is provided in Webb and Jarrett (2002). Peak flow data were obtained from Jarrett (2014) – Applied Weather Associates, Yochum and Moore (2013) – NRCS, from Kimbrough (2014) – USGS, and from the USGS’s National Water Information System (NWIS, http://waterdata.usgs.gov/co/nwis/sw). The majority of the estimates (NRCS, Jarrett) were developed using the critical depth method, in higher-gradient channels (> 1 percent). (The USGS primarily relied upon the slope-conveyance and step-backwater methods.) Using the critical-depth method, peak flow estimates are made using a single cross section, implementing high water marks at each location. Replicate measurements (separate flow estimates developed for several adjacent cross sections) were made for each reach of interest, to assess the reliability of the overall estimate. During floods, higher-gradient channels can flow at or near critical depth, where the Froude number (Fr) is unity and the following equation is applied to each cross section: V Fr 1 == (1) gD where, V is the average cross section velocity, D is the average flow depth (D = A/Tw), A is the 3rdJFIC, 2015 539 10thFISC+5thFIHMC PROCEEDINGS of the 3rd Joint Federal Interagency Conference on Sedimentation and Hydrologic Modeling, April 19-23, 2015, Reno, Nevada, USA flow area, Tw is the top width, and g is the acceleration due to gravity. Using the continuity equation, Q = VA, the Froude number equation can be reformulated to obtain flow rate: gA = AQ (2) Tw where Q is the discharge in cubic feet per second (cfs). Flood-Frequency Analysis The flood frequency analyses at streamgages where 2013 flood peaks are available were performed using logPearson frequency analyses as described in Bulletin 17B (Interagency Advisory Committee on Water Data, 1982), with the assumptions of independence and stationarity. The Expected Moments Algorithm (EMA) method, to be presented in Bulletin 17C, was not implemented. 2013 peak flows were included in the analyses. Record lengths varied from 131 to 10 years, with an average of 45 years. All of these streamgage records are affected by flow regulation (stream diversions and reservoirs). Weighted generalized skews were implemented for streamgages that had a sufficient number of similar watersheds for grouping, to adjust results for stream gages with a shorter record length. Flagged outliers were typically retained, unless they were confirmed as high outliers associated with significant dam failures. The historic peak algorithm was implemented for records that included historic peaks. Peak flow values for the remainder of the periods of record were obtained from the USGS NWIS system (http://waterdata.usgs.gov/co/nwis/sw), and the Colorado Division of Water Resources (http://www.dwr.state.co.us/Surfacewater/default.aspx). RESULTS AND DISCUSSION Peak flow estimates of up to 60,000 cfs (S. Platte River at Fort Morgan) were computed (Appendix A), with the highest discharges measured in the St. Vrain, Left Hand, Boulder, Little Thompson, and Big Thompson watersheds (Figure 1, Table 2), as well as the S. Platte River into which all of these streams drain. Detailed figures illustrating peak flow measurement points, as well as unit discharges and return intervals, are provided (Figures 3 through 6). In some of these streams, this peak flow was the flood of record (St. Vrain, 122 years of record; Left Hand, 17 years; Little Thompson, 18 years), while other impacted streams have higher flows in their streamgage records (Big Thompson River, 1976; Boulder Creek, 1921; S. Platte River, 1935). Table 1 Peak flow unit discharges by watershed size. Watshed Size (mi2) < 4 4 to 25 25 to 200 > 200 2 Maximum Unit Discharge (cfs/mi ) 1340 480 320 114 Unit discharges of up to 1340 cfs/mi2 were computed, with these unit discharges varying by watershed size (Table 1). The variation by scale illustrates the variation in precipitation depth and intensity by area, with some local areas, such as mountain slopes, receiving substantially more rainfall than other areas through such a mechanism as orographic forcing.